Upload folder using huggingface_hub

adapter_config.json

@@ -30,13 +30,13 @@
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
-    "q_proj",
+    "gate_proj",
     "k_proj",
     "up_proj",
     "down_proj",
-    "v_proj",
+    "q_proj",
     "o_proj",
-    "gate_proj"
+    "v_proj"
   ],
   "target_parameters": null,
   "task_type": "CAUSAL_LM",

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d948e97ba9490fe78e6232c27bfb6d7fa48ca4c6e76a7b4d14eb15b5abe8832d
+oid sha256:8e8b9d18a741945d57ae9c50da19b27c7eae565676488b34040cc2ed3b73a710
 size 36981856

optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:cd7e2258180cb3e934c4452f437ea0866f29adc4f3ef52c501fc3b271f6cab99
+oid sha256:6b34f6db49d513fcbf099dad7bf2cc4bb5d2188e1757be60413d2baa22cd7f39
 size 74194619

ref/adapter_config.json

@@ -30,13 +30,13 @@
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
-    "q_proj",
+    "gate_proj",
     "k_proj",
     "up_proj",
     "down_proj",
-    "v_proj",
+    "q_proj",
     "o_proj",
-    "gate_proj"
+    "v_proj"
   ],
   "target_parameters": null,
   "task_type": "CAUSAL_LM",

ref/adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c3f9d94657ed668f636e0c983388a97e23386ca701b294b2f543671a98c6ace9
+oid sha256:2bd76acdb9568aa0eaa808e5fdf632aef392350216ede4f4af8e7daeb17647b8
 size 73911112

rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:01f9a0f7843a37be87edd23f4e88aa93b38b95cc2c07503eeb1cf2e4632453a2
+oid sha256:f4a9f217e852f439efa6bd32fde98d6867f11aa6ea13ddc021ba10af6a0b0934
 size 14645

trainer_state.json

@@ -2,7 +2,7 @@
   "best_global_step": null,
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 6.0,
+  "epoch": 3.0,
   "eval_steps": 500,
   "global_step": 60,
   "is_hyper_param_search": false,
@@ -10,17 +10,17 @@
   "is_world_process_zero": true,
   "log_history": [
     {
-      "entropy": 1.512537270784378,
-      "epoch": 0.1038961038961039,
-      "grad_norm": 3.953125,
+      "entropy": 1.6346618384122849,
+      "epoch": 0.050314465408805034,
+      "grad_norm": 3.6875,
       "learning_rate": 5e-05,
-      "logits/chosen": -0.8704199966929909,
-      "logits/rejected": -0.3084566421487745,
-      "logps/chosen": -135.72114944458008,
-      "logps/rejected": -185.16453552246094,
+      "logits/chosen": -0.7411658949470903,
+      "logits/rejected": -0.20507352810005436,
+      "logps/chosen": -152.59497356414795,
+      "logps/rejected": -195.90787506103516,
       "loss": 0.6931471824645996,
-      "mean_token_accuracy": 0.5898260846734047,
-      "num_tokens": 7838.0,
+      "mean_token_accuracy": 0.562163695693016,
+      "num_tokens": 6930.0,
       "rewards/accuracies": 0.0,
       "rewards/chosen": 0.0,
       "rewards/margins": 0.0,
@@ -28,1072 +28,1072 @@
       "step": 1
     },
     {
-      "entropy": 1.5321979224681854,
-      "epoch": 0.2077922077922078,
-      "grad_norm": 3.203125,
+      "entropy": 1.4591333270072937,
+      "epoch": 0.10062893081761007,
+      "grad_norm": 3.375,
       "learning_rate": 4.9166666666666665e-05,
-      "logits/chosen": -1.1463743353817795,
-      "logits/rejected": -0.46985809672499795,
-      "logps/chosen": -128.28354263305664,
-      "logps/rejected": -196.00814056396484,
-      "loss": 0.5517404079437256,
-      "mean_token_accuracy": 0.5888336226344109,
-      "num_tokens": 15415.0,
+      "logits/chosen": -0.9232633579440521,
+      "logits/rejected": -0.42599087972727157,
+      "logps/chosen": -153.28020477294922,
+      "logps/rejected": -180.01216506958008,
+      "loss": 0.5030958652496338,
+      "mean_token_accuracy": 0.583746887743473,
+      "num_tokens": 12841.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.07305936858756468,
-      "rewards/margins": 0.3098052069544792,
-      "rewards/rejected": -0.23674583807587624,
+      "rewards/chosen": 0.22939472552388906,
+      "rewards/margins": 0.42953281849622726,
+      "rewards/rejected": -0.20013809762895107,
       "step": 2
     },
     {
-      "entropy": 1.5711090564727783,
-      "epoch": 0.3116883116883117,
-      "grad_norm": 2.6875,
+      "entropy": 1.4469595104455948,
+      "epoch": 0.1509433962264151,
+      "grad_norm": 2.59375,
       "learning_rate": 4.8333333333333334e-05,
-      "logits/chosen": -0.9020447850475974,
-      "logits/rejected": -0.36287321495179026,
-      "logps/chosen": -158.0212106704712,
-      "logps/rejected": -212.57695960998535,
-      "loss": 0.44210872054100037,
-      "mean_token_accuracy": 0.5543163530528545,
-      "num_tokens": 23608.0,
+      "logits/chosen": -1.0695778754695249,
+      "logits/rejected": -0.5926211969909361,
+      "logps/chosen": -146.45265579223633,
+      "logps/rejected": -187.52983283996582,
+      "loss": 0.38490599393844604,
+      "mean_token_accuracy": 0.6039908826351166,
+      "num_tokens": 18816.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.08782930835150182,
-      "rewards/margins": 0.6035905033349991,
-      "rewards/rejected": -0.5157611928880215,
+      "rewards/chosen": 0.34354821499437094,
+      "rewards/margins": 0.7708233185112476,
+      "rewards/rejected": -0.4272750951349735,
       "step": 3
     },
     {
-      "entropy": 1.4402711540460587,
-      "epoch": 0.4155844155844156,
-      "grad_norm": 2.109375,
+      "entropy": 1.518604964017868,
+      "epoch": 0.20125786163522014,
+      "grad_norm": 2.140625,
       "learning_rate": 4.75e-05,
-      "logits/chosen": -1.2677209306160548,
-      "logits/rejected": -0.4948322601683077,
-      "logps/chosen": -129.5712432861328,
-      "logps/rejected": -207.05406188964844,
-      "loss": 0.27849292755126953,
-      "mean_token_accuracy": 0.6066802740097046,
-      "num_tokens": 31225.0,
+      "logits/chosen": -0.935088641030341,
+      "logits/rejected": -0.3929027095055847,
+      "logps/chosen": -118.26776504516602,
+      "logps/rejected": -181.97826480865479,
+      "loss": 0.34729430079460144,
+      "mean_token_accuracy": 0.5912635922431946,
+      "num_tokens": 25522.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.31562281399965286,
-      "rewards/margins": 1.1901824250817299,
-      "rewards/rejected": -0.8745595999062061,
+      "rewards/chosen": 0.3668609778396785,
+      "rewards/margins": 0.9369952641427517,
+      "rewards/rejected": -0.5701342783868313,
       "step": 4
     },
     {
-      "entropy": 1.4334967136383057,
-      "epoch": 0.5194805194805194,
-      "grad_norm": 1.8984375,
+      "entropy": 1.436258003115654,
+      "epoch": 0.25157232704402516,
+      "grad_norm": 1.7578125,
       "learning_rate": 4.666666666666667e-05,
-      "logits/chosen": -1.1399579628249235,
-      "logits/rejected": -0.5072241096172478,
-      "logps/chosen": -129.72084999084473,
-      "logps/rejected": -186.66267013549805,
-      "loss": 0.27793484926223755,
-      "mean_token_accuracy": 0.5961734727025032,
-      "num_tokens": 38819.0,
+      "logits/chosen": -1.230789349374424,
+      "logits/rejected": -0.6072302734593342,
+      "logps/chosen": -118.5812578201294,
+      "logps/rejected": -193.36974906921387,
+      "loss": 0.21708804368972778,
+      "mean_token_accuracy": 0.5765212289988995,
+      "num_tokens": 32576.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.1814193739555776,
-      "rewards/margins": 1.1725517585873604,
-      "rewards/rejected": -0.9911323636770248,
+      "rewards/chosen": 0.4549223016947508,
+      "rewards/margins": 1.485500916838646,
+      "rewards/rejected": -1.0305786281824112,
       "step": 5
     },
     {
-      "entropy": 1.397928535938263,
-      "epoch": 0.6233766233766234,
-      "grad_norm": 1.3515625,
+      "entropy": 1.4615814238786697,
+      "epoch": 0.3018867924528302,
+      "grad_norm": 1.515625,
       "learning_rate": 4.5833333333333334e-05,
-      "logits/chosen": -1.2421275742450641,
-      "logits/rejected": -0.5668832705948026,
-      "logps/chosen": -140.5345287322998,
-      "logps/rejected": -216.73953437805176,
-      "loss": 0.16265499591827393,
-      "mean_token_accuracy": 0.6151033267378807,
-      "num_tokens": 46735.0,
+      "logits/chosen": -1.2629337189976264,
+      "logits/rejected": -0.4917354064790751,
+      "logps/chosen": -120.08106422424316,
+      "logps/rejected": -203.18199729919434,
+      "loss": 0.1837618350982666,
+      "mean_token_accuracy": 0.5985999256372452,
+      "num_tokens": 39703.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.21190567314624786,
-      "rewards/margins": 1.9610275849699974,
-      "rewards/rejected": -1.7491218969225883,
+      "rewards/chosen": 0.42343464493751526,
+      "rewards/margins": 1.681808888912201,
+      "rewards/rejected": -1.2583742290735245,
       "step": 6
     },
     {
-      "entropy": 1.3962671607732773,
-      "epoch": 0.7272727272727273,
-      "grad_norm": 0.87109375,
+      "entropy": 1.303115427494049,
+      "epoch": 0.3522012578616352,
+      "grad_norm": 1.5703125,
       "learning_rate": 4.5e-05,
-      "logits/chosen": -1.3785747536026947,
-      "logits/rejected": -0.7622838923063787,
-      "logps/chosen": -149.59703826904297,
-      "logps/rejected": -215.6246738433838,
-      "loss": 0.0853467807173729,
-      "mean_token_accuracy": 0.5751035511493683,
-      "num_tokens": 54606.0,
+      "logits/chosen": -1.3031889189529362,
+      "logits/rejected": -0.5977883441101891,
+      "logps/chosen": -108.28875637054443,
+      "logps/rejected": -193.97905731201172,
+      "loss": 0.1922958791255951,
+      "mean_token_accuracy": 0.6349928826093674,
+      "num_tokens": 47481.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.3336143558844924,
-      "rewards/margins": 2.5562889873981476,
-      "rewards/rejected": -2.222674608230591,
+      "rewards/chosen": 0.26323529286310077,
+      "rewards/margins": 1.6920193135738373,
+      "rewards/rejected": -1.428784005343914,
       "step": 7
     },
     {
-      "entropy": 1.3736034631729126,
-      "epoch": 0.8311688311688312,
-      "grad_norm": 1.078125,
+      "entropy": 1.3591983914375305,
+      "epoch": 0.4025157232704403,
+      "grad_norm": 1.0234375,
       "learning_rate": 4.4166666666666665e-05,
-      "logits/chosen": -1.5070403658179665,
-      "logits/rejected": -0.5901145721115999,
-      "logps/chosen": -135.39935779571533,
-      "logps/rejected": -214.47946739196777,
-      "loss": 0.10686549544334412,
-      "mean_token_accuracy": 0.583378367125988,
-      "num_tokens": 62397.0,
+      "logits/chosen": -1.42433679119847,
+      "logits/rejected": -0.987506627964361,
+      "logps/chosen": -137.10870265960693,
+      "logps/rejected": -202.67957878112793,
+      "loss": 0.10500691831111908,
+      "mean_token_accuracy": 0.6057217642664909,
+      "num_tokens": 54551.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.11785926297307014,
-      "rewards/margins": 2.4758750796318054,
-      "rewards/rejected": -2.3580158352851868,
+      "rewards/chosen": 0.6465687770396471,
+      "rewards/margins": 2.516775831580162,
+      "rewards/rejected": -1.87020705640316,
       "step": 8
     },
     {
-      "entropy": 1.2798690795898438,
-      "epoch": 0.935064935064935,
-      "grad_norm": 0.474609375,
+      "entropy": 1.305082157254219,
+      "epoch": 0.4528301886792453,
+      "grad_norm": 0.70703125,
       "learning_rate": 4.3333333333333334e-05,
-      "logits/chosen": -1.5770302287927562,
-      "logits/rejected": -1.0807285246276348,
-      "logps/chosen": -143.6166696548462,
-      "logps/rejected": -231.476957321167,
-      "loss": 0.04057425633072853,
-      "mean_token_accuracy": 0.6233189776539803,
-      "num_tokens": 70570.0,
+      "logits/chosen": -1.4197879173426369,
+      "logits/rejected": -0.9261144061875269,
+      "logps/chosen": -133.74537086486816,
+      "logps/rejected": -214.5505771636963,
+      "loss": 0.07105977088212967,
+      "mean_token_accuracy": 0.6148578971624374,
+      "num_tokens": 61082.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.40784178394824266,
-      "rewards/margins": 3.4395263493061066,
-      "rewards/rejected": -3.0316845774650574,
+      "rewards/chosen": 0.7314652749337256,
+      "rewards/margins": 3.20961757004261,
+      "rewards/rejected": -2.4781522899866104,
       "step": 9
     },
     {
-      "entropy": 1.3829500436782838,
-      "epoch": 1.0,
-      "grad_norm": 1.0234375,
+      "entropy": 1.3140722662210464,
+      "epoch": 0.5031446540880503,
+      "grad_norm": 0.54296875,
       "learning_rate": 4.25e-05,
-      "logits/chosen": -1.6479851532264296,
-      "logits/rejected": -0.9067774452457833,
-      "logps/chosen": -151.33749084472657,
-      "logps/rejected": -232.5306884765625,
-      "loss": 0.08217030763626099,
-      "mean_token_accuracy": 0.6009067416191101,
-      "num_tokens": 75516.0,
+      "logits/chosen": -1.7090569959939388,
+      "logits/rejected": -1.1499504201609827,
+      "logps/chosen": -114.76737594604492,
+      "logps/rejected": -215.3073024749756,
+      "loss": 0.04022533446550369,
+      "mean_token_accuracy": 0.5977272689342499,
+      "num_tokens": 68063.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": -0.06317704170942307,
-      "rewards/margins": 2.9441545248031615,
-      "rewards/rejected": -3.007331609725952,
+      "rewards/chosen": 0.7901867162436247,
+      "rewards/margins": 3.9837906509637833,
+      "rewards/rejected": -3.1936039328575134,
       "step": 10
     },
     {
-      "entropy": 1.3026847392320633,
-      "epoch": 1.103896103896104,
-      "grad_norm": 0.2158203125,
+      "entropy": 1.2443669438362122,
+      "epoch": 0.5534591194968553,
+      "grad_norm": 0.2734375,
       "learning_rate": 4.166666666666667e-05,
-      "logits/chosen": -1.7204814031981435,
-      "logits/rejected": -1.1328579956673597,
-      "logps/chosen": -164.92389488220215,
-      "logps/rejected": -254.53803253173828,
-      "loss": 0.01344618946313858,
-      "mean_token_accuracy": 0.5737045928835869,
-      "num_tokens": 83403.0,
+      "logits/chosen": -1.6404496463752554,
+      "logits/rejected": -1.1740173366915783,
+      "logps/chosen": -136.84611701965332,
+      "logps/rejected": -222.02164459228516,
+      "loss": 0.02123747020959854,
+      "mean_token_accuracy": 0.6184160858392715,
+      "num_tokens": 73558.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.21888723596930504,
-      "rewards/margins": 5.129873305559158,
-      "rewards/rejected": -4.910986036062241,
+      "rewards/chosen": 1.1639278389047831,
+      "rewards/margins": 4.731138825416565,
+      "rewards/rejected": -3.567211002111435,
       "step": 11
     },
     {
-      "entropy": 1.2513943165540695,
-      "epoch": 1.2077922077922079,
-      "grad_norm": 0.1982421875,
+      "entropy": 1.2719001024961472,
+      "epoch": 0.6037735849056604,
+      "grad_norm": 0.177734375,
       "learning_rate": 4.0833333333333334e-05,
-      "logits/chosen": -2.019919467615345,
-      "logits/rejected": -1.371831380074324,
-      "logps/chosen": -136.46048736572266,
-      "logps/rejected": -227.8621711730957,
-      "loss": 0.014287861995398998,
-      "mean_token_accuracy": 0.6023998335003853,
-      "num_tokens": 91399.0,
+      "logits/chosen": -1.7234823190264654,
+      "logits/rejected": -1.2490433686033002,
+      "logps/chosen": -133.34080696105957,
+      "logps/rejected": -222.08237648010254,
+      "loss": 0.01416418794542551,
+      "mean_token_accuracy": 0.6046521663665771,
+      "num_tokens": 79421.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.47683279449120164,
-      "rewards/margins": 4.8368901014328,
-      "rewards/rejected": -4.360057294368744,
+      "rewards/chosen": 1.091646859422326,
+      "rewards/margins": 4.951511919498444,
+      "rewards/rejected": -3.8598650693893433,
       "step": 12
     },
     {
-      "entropy": 1.2212176620960236,
-      "epoch": 1.3116883116883118,
-      "grad_norm": 0.2119140625,
+      "entropy": 1.2260268032550812,
+      "epoch": 0.6540880503144654,
+      "grad_norm": 0.4765625,
       "learning_rate": 4e-05,
-      "logits/chosen": -1.7292757979585844,
-      "logits/rejected": -1.1263007681151962,
-      "logps/chosen": -116.3019323348999,
-      "logps/rejected": -227.41853141784668,
-      "loss": 0.015741553157567978,
-      "mean_token_accuracy": 0.6425232589244843,
-      "num_tokens": 99387.0,
+      "logits/chosen": -1.8490552369674467,
+      "logits/rejected": -1.3316055711821662,
+      "logps/chosen": -130.4850263595581,
+      "logps/rejected": -238.74424743652344,
+      "loss": 0.031002508476376534,
+      "mean_token_accuracy": 0.623911663889885,
+      "num_tokens": 86682.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.29560147132724524,
-      "rewards/margins": 4.693124741315842,
-      "rewards/rejected": -4.39752334356308,
+      "rewards/chosen": 0.5156277692876756,
+      "rewards/margins": 5.114428728818893,
+      "rewards/rejected": -4.598800882697105,
       "step": 13
     },
     {
-      "entropy": 1.2582492232322693,
-      "epoch": 1.4155844155844157,
-      "grad_norm": 0.039794921875,
+      "entropy": 1.1846471428871155,
+      "epoch": 0.7044025157232704,
+      "grad_norm": 0.2734375,
       "learning_rate": 3.9166666666666665e-05,
-      "logits/chosen": -1.8986988604942008,
-      "logits/rejected": -1.2704682493398296,
-      "logps/chosen": -130.53850650787354,
-      "logps/rejected": -259.06455993652344,
-      "loss": 0.002613399177789688,
-      "mean_token_accuracy": 0.6116124093532562,
-      "num_tokens": 106985.0,
+      "logits/chosen": -1.6988998158408573,
+      "logits/rejected": -1.412297648550121,
+      "logps/chosen": -134.32691383361816,
+      "logps/rejected": -220.49063682556152,
+      "loss": 0.021130766719579697,
+      "mean_token_accuracy": 0.6293000131845474,
+      "num_tokens": 93035.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.37937870621681213,
-      "rewards/margins": 6.370364427566528,
-      "rewards/rejected": -5.990985751152039,
+      "rewards/chosen": 0.28157730400562286,
+      "rewards/margins": 4.996448755264282,
+      "rewards/rejected": -4.714871436357498,
       "step": 14
     },
     {
-      "entropy": 1.2522720396518707,
-      "epoch": 1.5194805194805194,
-      "grad_norm": 0.05810546875,
+      "entropy": 1.1781003028154373,
+      "epoch": 0.7547169811320755,
+      "grad_norm": 0.0517578125,
       "learning_rate": 3.8333333333333334e-05,
-      "logits/chosen": -2.005344554741843,
-      "logits/rejected": -1.3727889740878687,
-      "logps/chosen": -155.52635955810547,
-      "logps/rejected": -286.3923683166504,
-      "loss": 0.002808060497045517,
-      "mean_token_accuracy": 0.5796493962407112,
-      "num_tokens": 114781.0,
+      "logits/chosen": -1.8902387361462176,
+      "logits/rejected": -1.4900272791975393,
+      "logps/chosen": -132.44914436340332,
+      "logps/rejected": -247.65433311462402,
+      "loss": 0.002454916015267372,
+      "mean_token_accuracy": 0.6452845111489296,
+      "num_tokens": 98572.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.3164886636659503,
-      "rewards/margins": 7.638198435306549,
-      "rewards/rejected": -7.3217098116874695,
+      "rewards/chosen": 1.2324292734265327,
+      "rewards/margins": 7.052622675895691,
+      "rewards/rejected": -5.820193350315094,
       "step": 15
     },
     {
-      "entropy": 1.2523068189620972,
-      "epoch": 1.6233766233766234,
-      "grad_norm": 0.038330078125,
+      "entropy": 1.2133459448814392,
+      "epoch": 0.8050314465408805,
+      "grad_norm": 0.20703125,
       "learning_rate": 3.7500000000000003e-05,
-      "logits/chosen": -1.9715578454090756,
-      "logits/rejected": -1.2781532380406784,
-      "logps/chosen": -150.2594985961914,
-      "logps/rejected": -280.48920822143555,
-      "loss": 0.002301866188645363,
-      "mean_token_accuracy": 0.6012703776359558,
-      "num_tokens": 122871.0,
+      "logits/chosen": -1.8413987197790387,
+      "logits/rejected": -1.4087605133817744,
+      "logps/chosen": -139.3100233078003,
+      "logps/rejected": -246.31277465820312,
+      "loss": 0.01590706966817379,
+      "mean_token_accuracy": 0.6142129823565483,
+      "num_tokens": 106014.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.19919815217144787,
-      "rewards/margins": 7.385057747364044,
-      "rewards/rejected": -7.1858596205711365,
+      "rewards/chosen": 0.3513742520008236,
+      "rewards/margins": 6.107073247432709,
+      "rewards/rejected": -5.755699068307877,
       "step": 16
     },
     {
-      "entropy": 1.234215423464775,
-      "epoch": 1.7272727272727273,
-      "grad_norm": 0.045654296875,
+      "entropy": 1.2062125205993652,
+      "epoch": 0.8553459119496856,
+      "grad_norm": 0.1396484375,
       "learning_rate": 3.6666666666666666e-05,
-      "logits/chosen": -1.9352984235246145,
-      "logits/rejected": -1.3910207936164074,
-      "logps/chosen": -142.13225173950195,
-      "logps/rejected": -268.2549476623535,
-      "loss": 0.002515692263841629,
-      "mean_token_accuracy": 0.5971521884202957,
-      "num_tokens": 130530.0,
+      "logits/chosen": -1.9615212533907105,
+      "logits/rejected": -1.3313654933874406,
+      "logps/chosen": -127.33739471435547,
+      "logps/rejected": -249.73816299438477,
+      "loss": 0.00578045891597867,
+      "mean_token_accuracy": 0.6195648014545441,
+      "num_tokens": 112297.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.42800588643876836,
-      "rewards/margins": 7.45609712600708,
-      "rewards/rejected": -7.028091222047806,
+      "rewards/chosen": 0.9851933747413568,
+      "rewards/margins": 7.103353261947632,
+      "rewards/rejected": -6.118159830570221,
       "step": 17
     },
     {
-      "entropy": 1.190779373049736,
-      "epoch": 1.8311688311688312,
-      "grad_norm": 0.0390625,
+      "entropy": 1.198799878358841,
+      "epoch": 0.9056603773584906,
+      "grad_norm": 0.030517578125,
       "learning_rate": 3.5833333333333335e-05,
-      "logits/chosen": -2.0085451937187595,
-      "logits/rejected": -1.6083773573241704,
-      "logps/chosen": -131.45305252075195,
-      "logps/rejected": -244.74785995483398,
-      "loss": 0.0026926116552203894,
-      "mean_token_accuracy": 0.5924705117940903,
-      "num_tokens": 138371.0,
+      "logits/chosen": -2.082073810115763,
+      "logits/rejected": -1.6705133301771156,
+      "logps/chosen": -128.168438911438,
+      "logps/rejected": -246.99494552612305,
+      "loss": 0.001798766665160656,
+      "mean_token_accuracy": 0.6067424267530441,
+      "num_tokens": 118259.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.27736073173582554,
-      "rewards/margins": 6.875377118587494,
-      "rewards/rejected": -6.598016321659088,
+      "rewards/chosen": 0.6913285218179226,
+      "rewards/margins": 7.232239246368408,
+      "rewards/rejected": -6.540910840034485,
       "step": 18
     },
     {
-      "entropy": 1.2075671404600143,
-      "epoch": 1.935064935064935,
-      "grad_norm": 0.0262451171875,
+      "entropy": 1.1391501128673553,
+      "epoch": 0.9559748427672956,
+      "grad_norm": 0.023193359375,
       "learning_rate": 3.5e-05,
-      "logits/chosen": -2.134337462790497,
-      "logits/rejected": -1.6570472501112965,
-      "logps/chosen": -120.83753395080566,
-      "logps/rejected": -264.36783027648926,
-      "loss": 0.001522539067082107,
-      "mean_token_accuracy": 0.5812797658145428,
-      "num_tokens": 146105.0,
+      "logits/chosen": -2.1592130772558074,
+      "logits/rejected": -1.7971208229649984,
+      "logps/chosen": -136.7722806930542,
+      "logps/rejected": -235.51027870178223,
+      "loss": 0.0014428942231461406,
+      "mean_token_accuracy": 0.6118658147752285,
+      "num_tokens": 123618.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.5302159509155899,
-      "rewards/margins": 7.605401933193207,
-      "rewards/rejected": -7.075186014175415,
+      "rewards/chosen": 1.2391198351979256,
+      "rewards/margins": 7.5219069719314575,
+      "rewards/rejected": -6.282787263393402,
       "step": 19
     },
     {
-      "entropy": 1.116961932182312,
-      "epoch": 2.0,
-      "grad_norm": 0.07470703125,
+      "entropy": 1.2272680827549525,
+      "epoch": 1.0,
+      "grad_norm": 0.037353515625,
       "learning_rate": 3.4166666666666666e-05,
-      "logits/chosen": -1.9841584232594507,
-      "logits/rejected": -1.6315209771433286,
-      "logps/chosen": -129.09716339111327,
-      "logps/rejected": -246.24557495117188,
-      "loss": 0.0038437608163803816,
-      "mean_token_accuracy": 0.6218155741691589,
-      "num_tokens": 151032.0,
+      "logits/chosen": -1.927751023508857,
+      "logits/rejected": -1.4586340540864533,
+      "logps/chosen": -140.12276676722936,
+      "logps/rejected": -256.6956220354353,
+      "loss": 0.0020336967427283525,
+      "mean_token_accuracy": 0.6004699298313686,
+      "num_tokens": 130000.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.2574792455881834,
-      "rewards/margins": 7.590111160278321,
-      "rewards/rejected": -7.332631969451905,
+      "rewards/chosen": -0.13402164301701955,
+      "rewards/margins": 6.993039471762521,
+      "rewards/rejected": -7.12706116267613,
       "step": 20
     },
     {
-      "entropy": 1.1428246572613716,
-      "epoch": 2.103896103896104,
-      "grad_norm": 0.046142578125,
+      "entropy": 1.1386558711528778,
+      "epoch": 1.050314465408805,
+      "grad_norm": 0.00396728515625,
       "learning_rate": 3.3333333333333335e-05,
-      "logits/chosen": -2.1280286073627903,
-      "logits/rejected": -1.6765060194642971,
-      "logps/chosen": -124.37408638000488,
-      "logps/rejected": -268.4387798309326,
-      "loss": 0.002784793497994542,
-      "mean_token_accuracy": 0.6262179091572762,
-      "num_tokens": 158898.0,
+      "logits/chosen": -2.222644827049755,
+      "logits/rejected": -1.7477538407871338,
+      "logps/chosen": -132.11598300933838,
+      "logps/rejected": -268.2633934020996,
+      "loss": 0.00020816404139623046,
+      "mean_token_accuracy": 0.6361734122037888,
+      "num_tokens": 135799.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.4877193462452851,
-      "rewards/margins": 8.688470482826233,
-      "rewards/rejected": -8.20075112581253,
+      "rewards/chosen": 1.087248451076448,
+      "rewards/margins": 8.924234092235565,
+      "rewards/rejected": -7.8369855880737305,
       "step": 21
     },
     {
-      "entropy": 1.1632115840911865,
-      "epoch": 2.207792207792208,
-      "grad_norm": 0.00421142578125,
+      "entropy": 1.188778430223465,
+      "epoch": 1.10062893081761,
+      "grad_norm": 0.06005859375,
       "learning_rate": 3.2500000000000004e-05,
-      "logits/chosen": -2.2110131627297314,
-      "logits/rejected": -1.624031390186785,
-      "logps/chosen": -143.07822036743164,
-      "logps/rejected": -304.6781883239746,
-      "loss": 0.00020159759151283652,
-      "mean_token_accuracy": 0.582977369427681,
-      "num_tokens": 166715.0,
+      "logits/chosen": -2.190930237663479,
+      "logits/rejected": -1.5339520761713392,
+      "logps/chosen": -136.57733154296875,
+      "logps/rejected": -276.55401611328125,
+      "loss": 0.0026830965653061867,
+      "mean_token_accuracy": 0.5980619117617607,
+      "num_tokens": 142686.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.3669938058592379,
-      "rewards/margins": 9.80722576379776,
-      "rewards/rejected": -9.440232157707214,
+      "rewards/chosen": 0.16770659349276684,
+      "rewards/margins": 8.182168662548065,
+      "rewards/rejected": -8.014461994171143,
       "step": 22
     },
     {
-      "entropy": 1.163627065718174,
-      "epoch": 2.311688311688312,
-      "grad_norm": 0.0189208984375,
+      "entropy": 1.1210491955280304,
+      "epoch": 1.150943396226415,
+      "grad_norm": 0.01141357421875,
       "learning_rate": 3.1666666666666666e-05,
-      "logits/chosen": -2.151204727301539,
-      "logits/rejected": -1.5880060075596425,
-      "logps/chosen": -161.0470733642578,
-      "logps/rejected": -283.40243339538574,
-      "loss": 0.0010750408982858062,
-      "mean_token_accuracy": 0.5792371965944767,
-      "num_tokens": 174871.0,
+      "logits/chosen": -2.2553734561327348,
+      "logits/rejected": -1.9669641015596422,
+      "logps/chosen": -121.1500768661499,
+      "logps/rejected": -249.81173515319824,
+      "loss": 0.0006848493940196931,
+      "mean_token_accuracy": 0.5800603404641151,
+      "num_tokens": 148888.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": -0.393167894333601,
-      "rewards/margins": 8.00706958770752,
-      "rewards/rejected": -8.400237500667572,
+      "rewards/chosen": 0.6980049340054393,
+      "rewards/margins": 8.123494029045105,
+      "rewards/rejected": -7.425489127635956,
       "step": 23
     },
     {
-      "entropy": 1.179183878004551,
-      "epoch": 2.4155844155844157,
-      "grad_norm": 0.01806640625,
+      "entropy": 1.2056218385696411,
+      "epoch": 1.20125786163522,
+      "grad_norm": 0.0213623046875,
       "learning_rate": 3.0833333333333335e-05,
-      "logits/chosen": -2.1281537439249174,
-      "logits/rejected": -1.5609932878048027,
-      "logps/chosen": -129.24316215515137,
-      "logps/rejected": -269.68304443359375,
-      "loss": 0.001012458698824048,
-      "mean_token_accuracy": 0.6229125708341599,
-      "num_tokens": 182581.0,
+      "logits/chosen": -2.0574360723332568,
+      "logits/rejected": -1.5004657659813077,
+      "logps/chosen": -141.36074924468994,
+      "logps/rejected": -285.9758243560791,
+      "loss": 0.0011920351535081863,
+      "mean_token_accuracy": 0.6025349199771881,
+      "num_tokens": 155066.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.30742750596255064,
-      "rewards/margins": 8.548649609088898,
-      "rewards/rejected": -8.241222083568573,
+      "rewards/chosen": 0.3790994770824909,
+      "rewards/margins": 9.149889826774597,
+      "rewards/rejected": -8.77079039812088,
       "step": 24
     },
     {
-      "entropy": 1.2349543124437332,
-      "epoch": 2.5194805194805197,
-      "grad_norm": 0.0118408203125,
+      "entropy": 1.0760537832975388,
+      "epoch": 1.251572327044025,
+      "grad_norm": 0.0164794921875,
       "learning_rate": 3e-05,
-      "logits/chosen": -2.1954152548130956,
-      "logits/rejected": -1.8124665579280124,
-      "logps/chosen": -139.44278526306152,
-      "logps/rejected": -276.38891220092773,
-      "loss": 0.000593633740209043,
-      "mean_token_accuracy": 0.5867471769452095,
-      "num_tokens": 190414.0,
+      "logits/chosen": -2.243240417111155,
+      "logits/rejected": -1.6917496542346284,
+      "logps/chosen": -123.56195449829102,
+      "logps/rejected": -266.39801025390625,
+      "loss": 0.0008477906812913716,
+      "mean_token_accuracy": 0.6462969928979874,
+      "num_tokens": 161896.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.11138467863202095,
-      "rewards/margins": 8.101492822170258,
-      "rewards/rejected": -7.990108132362366,
+      "rewards/chosen": 0.5940647050738335,
+      "rewards/margins": 8.513349890708923,
+      "rewards/rejected": -7.919285297393799,
       "step": 25
     },
     {
-      "entropy": 1.1430687606334686,
-      "epoch": 2.6233766233766236,
-      "grad_norm": 0.0186767578125,
+      "entropy": 1.1326584815979004,
+      "epoch": 1.3018867924528301,
+      "grad_norm": 0.0177001953125,
       "learning_rate": 2.916666666666667e-05,
-      "logits/chosen": -2.184405102771366,
-      "logits/rejected": -1.768430591565442,
-      "logps/chosen": -162.35870456695557,
-      "logps/rejected": -284.31261253356934,
-      "loss": 0.001006356906145811,
-      "mean_token_accuracy": 0.5789020732045174,
-      "num_tokens": 198352.0,
+      "logits/chosen": -2.025026339871026,
+      "logits/rejected": -1.6378500703330563,
+      "logps/chosen": -120.81675624847412,
+      "logps/rejected": -253.07993507385254,
+      "loss": 0.0010006717639043927,
+      "mean_token_accuracy": 0.6229780465364456,
+      "num_tokens": 168755.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": -0.32843170780688524,
-      "rewards/margins": 8.667101621627808,
-      "rewards/rejected": -8.995533287525177,
+      "rewards/chosen": 0.3348179815802723,
+      "rewards/margins": 7.985908567905426,
+      "rewards/rejected": -7.651090621948242,
       "step": 26
     },
     {
-      "entropy": 1.1707737743854523,
-      "epoch": 2.7272727272727275,
-      "grad_norm": 0.00994873046875,
+      "entropy": 1.1113788187503815,
+      "epoch": 1.3522012578616351,
+      "grad_norm": 0.00836181640625,
       "learning_rate": 2.8333333333333335e-05,
-      "logits/chosen": -2.2829095936666026,
-      "logits/rejected": -1.5454627881242438,
-      "logps/chosen": -127.9923505783081,
-      "logps/rejected": -269.86738777160645,
-      "loss": 0.000537040235940367,
-      "mean_token_accuracy": 0.5975877344608307,
-      "num_tokens": 206134.0,
+      "logits/chosen": -2.2742666534990894,
+      "logits/rejected": -1.9221175464707185,
+      "logps/chosen": -143.62575721740723,
+      "logps/rejected": -270.15936851501465,
+      "loss": 0.00037246744614094496,
+      "mean_token_accuracy": 0.6221201568841934,
+      "num_tokens": 175381.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": -0.15700979018583894,
-      "rewards/margins": 8.28273743391037,
-      "rewards/rejected": -8.439747273921967,
+      "rewards/chosen": 1.0057550426572561,
+      "rewards/margins": 9.40854561328888,
+      "rewards/rejected": -8.402790665626526,
       "step": 27
     },
     {
-      "entropy": 1.1310725212097168,
-      "epoch": 2.8311688311688314,
-      "grad_norm": 0.00567626953125,
+      "entropy": 1.1610162109136581,
+      "epoch": 1.4025157232704402,
+      "grad_norm": 0.00518798828125,
       "learning_rate": 2.7500000000000004e-05,
-      "logits/chosen": -2.319415917931936,
-      "logits/rejected": -1.8342084919918709,
-      "logps/chosen": -150.00816440582275,
-      "logps/rejected": -293.56144523620605,
-      "loss": 0.000298125552944839,
-      "mean_token_accuracy": 0.6064256802201271,
-      "num_tokens": 214297.0,
+      "logits/chosen": -2.122237574979544,
+      "logits/rejected": -1.8119841137964543,
+      "logps/chosen": -124.42662048339844,
+      "logps/rejected": -272.423641204834,
+      "loss": 0.0002692067064344883,
+      "mean_token_accuracy": 0.6206282004714012,
+      "num_tokens": 181771.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": -0.005369566380977631,
-      "rewards/margins": 9.238496720790863,
-      "rewards/rejected": -9.243866205215454,
+      "rewards/chosen": 0.9137327093631029,
+      "rewards/margins": 9.385262191295624,
+      "rewards/rejected": -8.47152954339981,
       "step": 28
     },
     {
-      "entropy": 1.1214349120855331,
-      "epoch": 2.935064935064935,
-      "grad_norm": 0.0010833740234375,
+      "entropy": 1.1231490969657898,
+      "epoch": 1.4528301886792452,
+      "grad_norm": 0.010009765625,
       "learning_rate": 2.6666666666666667e-05,
-      "logits/chosen": -2.39867742884011,
-      "logits/rejected": -1.938792760577064,
-      "logps/chosen": -129.17547798156738,
-      "logps/rejected": -291.3078727722168,
-      "loss": 5.1582956075435504e-05,
-      "mean_token_accuracy": 0.5967546105384827,
-      "num_tokens": 221900.0,
+      "logits/chosen": -2.3010927487353845,
+      "logits/rejected": -1.9966991698314587,
+      "logps/chosen": -139.3568572998047,
+      "logps/rejected": -258.50311279296875,
+      "loss": 0.0005843475228175521,
+      "mean_token_accuracy": 0.6224977001547813,
+      "num_tokens": 188644.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.385247141122818,
-      "rewards/margins": 11.02652394771576,
-      "rewards/rejected": -10.641276895999908,
+      "rewards/chosen": -0.0714789517223835,
+      "rewards/margins": 8.368709564208984,
+      "rewards/rejected": -8.440188646316528,
       "step": 29
     },
     {
-      "entropy": 1.1968913316726684,
-      "epoch": 3.0,
-      "grad_norm": 0.000522613525390625,
+      "entropy": 1.1115453839302063,
+      "epoch": 1.5031446540880502,
+      "grad_norm": 0.0267333984375,
       "learning_rate": 2.5833333333333336e-05,
-      "logits/chosen": -2.3927515248399525,
-      "logits/rejected": -1.7979455730124954,
-      "logps/chosen": -134.36846923828125,
-      "logps/rejected": -315.8888793945313,
-      "loss": 2.0089810277568176e-05,
-      "mean_token_accuracy": 0.5749531865119935,
-      "num_tokens": 226548.0,
+      "logits/chosen": -2.2130385961990293,
+      "logits/rejected": -1.8950639322000826,
+      "logps/chosen": -119.8697099685669,
+      "logps/rejected": -246.79789543151855,
+      "loss": 0.0016802519094198942,
+      "mean_token_accuracy": 0.6181407496333122,
+      "num_tokens": 196391.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.5599890232086182,
-      "rewards/margins": 11.858483695983887,
-      "rewards/rejected": -11.298494720458985,
+      "rewards/chosen": 0.025272462284192443,
+      "rewards/margins": 7.992879867553711,
+      "rewards/rejected": -7.967607498168945,
       "step": 30
     },
     {
-      "entropy": 1.0848063379526138,
-      "epoch": 3.103896103896104,
-      "grad_norm": 0.00848388671875,
+      "entropy": 1.086497575044632,
+      "epoch": 1.5534591194968552,
+      "grad_norm": 0.03271484375,
       "learning_rate": 2.5e-05,
-      "logits/chosen": -2.3630775053643998,
-      "logits/rejected": -2.0637178830428264,
-      "logps/chosen": -136.13608932495117,
-      "logps/rejected": -278.8615074157715,
-      "loss": 0.00036562219611369073,
-      "mean_token_accuracy": 0.628660149872303,
-      "num_tokens": 234538.0,
+      "logits/chosen": -2.2244215972877353,
+      "logits/rejected": -1.7872997006786144,
+      "logps/chosen": -159.92728996276855,
+      "logps/rejected": -284.250545501709,
+      "loss": 0.0013625127030536532,
+      "mean_token_accuracy": 0.6142807975411415,
+      "num_tokens": 203518.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.31311475299298763,
-      "rewards/margins": 9.501745641231537,
-      "rewards/rejected": -9.188630878925323,
+      "rewards/chosen": -0.06530415918678045,
+      "rewards/margins": 8.824142456054688,
+      "rewards/rejected": -8.88944661617279,
       "step": 31
     },
     {
-      "entropy": 1.162983387708664,
-      "epoch": 3.207792207792208,
-      "grad_norm": 0.00115203857421875,
+      "entropy": 1.136437065899372,
+      "epoch": 1.6037735849056602,
+      "grad_norm": 0.0081787109375,
       "learning_rate": 2.4166666666666667e-05,
-      "logits/chosen": -2.406029792298791,
-      "logits/rejected": -1.9060521152655108,
-      "logps/chosen": -126.35858535766602,
-      "logps/rejected": -303.76181411743164,
-      "loss": 5.3637420933227986e-05,
-      "mean_token_accuracy": 0.5992156118154526,
-      "num_tokens": 242224.0,
+      "logits/chosen": -2.211966660335218,
+      "logits/rejected": -1.7966088455890876,
+      "logps/chosen": -143.01754474639893,
+      "logps/rejected": -294.5742607116699,
+      "loss": 0.0003461150045040995,
+      "mean_token_accuracy": 0.6098548360168934,
+      "num_tokens": 209850.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.3918016329407692,
-      "rewards/margins": 10.524143934249878,
-      "rewards/rejected": -10.132342457771301,
+      "rewards/chosen": 0.2523078881204128,
+      "rewards/margins": 10.28915536403656,
+      "rewards/rejected": -10.036847531795502,
       "step": 32
     },
     {
-      "entropy": 1.1035910472273827,
-      "epoch": 3.311688311688312,
-      "grad_norm": 0.00885009765625,
+      "entropy": 1.130326434969902,
+      "epoch": 1.6540880503144653,
+      "grad_norm": 0.01904296875,
       "learning_rate": 2.3333333333333336e-05,
-      "logits/chosen": -2.4029672208683124,
-      "logits/rejected": -1.8559939119150626,
-      "logps/chosen": -164.47723293304443,
-      "logps/rejected": -307.7430419921875,
-      "loss": 0.0004137605137657374,
-      "mean_token_accuracy": 0.5645543709397316,
-      "num_tokens": 250221.0,
+      "logits/chosen": -2.4035100083486487,
+      "logits/rejected": -1.7515896245587153,
+      "logps/chosen": -125.12364768981934,
+      "logps/rejected": -263.9906826019287,
+      "loss": 0.0005816172924824059,
+      "mean_token_accuracy": 0.6256594210863113,
+      "num_tokens": 215637.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": -0.5422927811741829,
-      "rewards/margins": 10.185687839984894,
-      "rewards/rejected": -10.72798067331314,
+      "rewards/chosen": 0.7825962388888001,
+      "rewards/margins": 9.555868089199066,
+      "rewards/rejected": -8.77327173948288,
       "step": 33
     },
     {
-      "entropy": 1.1141554564237595,
-      "epoch": 3.4155844155844157,
-      "grad_norm": 0.007110595703125,
+      "entropy": 1.0617773309350014,
+      "epoch": 1.7044025157232703,
+      "grad_norm": 0.0081787109375,
       "learning_rate": 2.25e-05,
-      "logits/chosen": -2.3687534580534138,
-      "logits/rejected": -2.050704117333944,
-      "logps/chosen": -138.01462364196777,
-      "logps/rejected": -276.9965763092041,
-      "loss": 0.0003196091565769166,
-      "mean_token_accuracy": 0.5838265940546989,
-      "num_tokens": 258004.0,
+      "logits/chosen": -2.450243334633306,
+      "logits/rejected": -2.0934474514571817,
+      "logps/chosen": -125.30186462402344,
+      "logps/rejected": -256.15345764160156,
+      "loss": 0.0004422894271556288,
+      "mean_token_accuracy": 0.6317535266280174,
+      "num_tokens": 221051.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.1458825245499611,
-      "rewards/margins": 10.098270773887634,
-      "rewards/rejected": -9.952388286590576,
+      "rewards/chosen": 1.3803571164608002,
+      "rewards/margins": 9.975262939929962,
+      "rewards/rejected": -8.594905972480774,
       "step": 34
     },
     {
-      "entropy": 1.2330412343144417,
-      "epoch": 3.5194805194805197,
-      "grad_norm": 0.0016021728515625,
+      "entropy": 1.1356039494276047,
+      "epoch": 1.7547169811320755,
+      "grad_norm": 0.0032958984375,
       "learning_rate": 2.1666666666666667e-05,
-      "logits/chosen": -2.099665713097579,
-      "logits/rejected": -1.4495526677555262,
-      "logps/chosen": -147.08791255950928,
-      "logps/rejected": -340.46545028686523,
-      "loss": 6.672517338301986e-05,
-      "mean_token_accuracy": 0.5911035016179085,
-      "num_tokens": 265864.0,
+      "logits/chosen": -2.364045478373609,
+      "logits/rejected": -2.100374343065235,
+      "logps/chosen": -131.08294582366943,
+      "logps/rejected": -274.32538986206055,
+      "loss": 0.00013096419570501894,
+      "mean_token_accuracy": 0.6208470910787582,
+      "num_tokens": 226426.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": -0.028002352453768253,
-      "rewards/margins": 11.662676930427551,
-      "rewards/rejected": -11.690679430961609,
+      "rewards/chosen": 0.9078209921717644,
+      "rewards/margins": 10.476734340190887,
+      "rewards/rejected": -9.568913280963898,
       "step": 35
     },
     {
-      "entropy": 1.1366761773824692,
-      "epoch": 3.6233766233766236,
-      "grad_norm": 0.02197265625,
+      "entropy": 1.1196807771921158,
+      "epoch": 1.8050314465408805,
+      "grad_norm": 0.0201416015625,
       "learning_rate": 2.0833333333333336e-05,
-      "logits/chosen": -2.328846034176437,
-      "logits/rejected": -1.992655699239898,
-      "logps/chosen": -135.61105346679688,
-      "logps/rejected": -271.5671138763428,
-      "loss": 0.0011770055862143636,
-      "mean_token_accuracy": 0.6069249212741852,
-      "num_tokens": 273581.0,
+      "logits/chosen": -2.4489547712147877,
+      "logits/rejected": -1.96198692605538,
+      "logps/chosen": -116.7791223526001,
+      "logps/rejected": -270.7232780456543,
+      "loss": 0.0009249980794265866,
+      "mean_token_accuracy": 0.623729519546032,
+      "num_tokens": 233419.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.13576746042235754,
-      "rewards/margins": 8.846820831298828,
-      "rewards/rejected": -8.711053371429443,
+      "rewards/chosen": 0.3964777053333819,
+      "rewards/margins": 9.61779636144638,
+      "rewards/rejected": -9.221318662166595,
       "step": 36
     },
     {
-      "entropy": 1.1373939514160156,
-      "epoch": 3.7272727272727275,
-      "grad_norm": 0.006072998046875,
+      "entropy": 1.0988834276795387,
+      "epoch": 1.8553459119496856,
+      "grad_norm": 0.002960205078125,
       "learning_rate": 2e-05,
-      "logits/chosen": -2.2702114795686574,
-      "logits/rejected": -1.7482615971180808,
-      "logps/chosen": -140.26164054870605,
-      "logps/rejected": -294.61451530456543,
-      "loss": 0.0003134559956379235,
-      "mean_token_accuracy": 0.5956408530473709,
-      "num_tokens": 281309.0,
+      "logits/chosen": -2.347542994631864,
+      "logits/rejected": -1.9687581260709852,
+      "logps/chosen": -144.3069133758545,
+      "logps/rejected": -294.52979278564453,
+      "loss": 0.0001527789863757789,
+      "mean_token_accuracy": 0.5927288830280304,
+      "num_tokens": 240673.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.032456397311761975,
-      "rewards/margins": 10.235666334629059,
-      "rewards/rejected": -10.203210055828094,
+      "rewards/chosen": -0.041503352113068104,
+      "rewards/margins": 10.00435733795166,
+      "rewards/rejected": -10.045860767364502,
       "step": 37
     },
     {
-      "entropy": 1.1416693180799484,
-      "epoch": 3.8311688311688314,
-      "grad_norm": 0.00628662109375,
+      "entropy": 1.09148870408535,
+      "epoch": 1.9056603773584906,
+      "grad_norm": 0.0035552978515625,
       "learning_rate": 1.9166666666666667e-05,
-      "logits/chosen": -2.324989094221912,
-      "logits/rejected": -1.6231121376823092,
-      "logps/chosen": -123.99202156066895,
-      "logps/rejected": -272.5278739929199,
-      "loss": 0.000329699891153723,
-      "mean_token_accuracy": 0.6103288903832436,
-      "num_tokens": 289131.0,
+      "logits/chosen": -2.319408068295181,
+      "logits/rejected": -2.07876476569641,
+      "logps/chosen": -140.95322799682617,
+      "logps/rejected": -283.8196105957031,
+      "loss": 0.0001463937369408086,
+      "mean_token_accuracy": 0.6046550124883652,
+      "num_tokens": 245974.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": -0.029055589344352484,
-      "rewards/margins": 8.933434963226318,
-      "rewards/rejected": -8.962490499019623,
+      "rewards/chosen": 0.7766995467245579,
+      "rewards/margins": 10.631062924861908,
+      "rewards/rejected": -9.85436338186264,
       "step": 38
     },
     {
-      "entropy": 1.1619316935539246,
-      "epoch": 3.935064935064935,
-      "grad_norm": 0.006805419921875,
+      "entropy": 1.1747687011957169,
+      "epoch": 1.9559748427672956,
+      "grad_norm": 0.00439453125,
       "learning_rate": 1.8333333333333333e-05,
-      "logits/chosen": -2.277468688630792,
-      "logits/rejected": -1.6969099471315172,
-      "logps/chosen": -139.51892566680908,
-      "logps/rejected": -279.6028060913086,
-      "loss": 0.00035776293952949345,
-      "mean_token_accuracy": 0.6005248352885246,
-      "num_tokens": 297124.0,
+      "logits/chosen": -2.1828008029912604,
+      "logits/rejected": -1.726034681820153,
+      "logps/chosen": -126.46953868865967,
+      "logps/rejected": -270.92620849609375,
+      "loss": 0.0002291620330652222,
+      "mean_token_accuracy": 0.6064154356718063,
+      "num_tokens": 253184.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": -0.09404934721533209,
-      "rewards/margins": 9.114178478717804,
-      "rewards/rejected": -9.208227813243866,
+      "rewards/chosen": -0.02468252368271351,
+      "rewards/margins": 9.081076204776764,
+      "rewards/rejected": -9.105758786201477,
       "step": 39
     },
     {
-      "entropy": 1.1786307096481323,
-      "epoch": 4.0,
-      "grad_norm": 0.003021240234375,
+      "entropy": 1.1650119509015764,
+      "epoch": 2.0,
+      "grad_norm": 0.02099609375,
       "learning_rate": 1.75e-05,
-      "logits/chosen": -2.4174767570194224,
-      "logits/rejected": -1.7008187103422685,
-      "logps/chosen": -175.12867736816406,
-      "logps/rejected": -322.8905456542969,
-      "loss": 0.00012227564002387226,
-      "mean_token_accuracy": 0.5624364495277405,
-      "num_tokens": 302064.0,
+      "logits/chosen": -2.2664670930140036,
+      "logits/rejected": -1.7445590165072045,
+      "logps/chosen": -153.27142769949776,
+      "logps/rejected": -289.8901563371931,
+      "loss": 0.0010365444468334317,
+      "mean_token_accuracy": 0.5644707764898028,
+      "num_tokens": 260000.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": -0.3554480280727148,
-      "rewards/margins": 10.46877145767212,
-      "rewards/rejected": -10.824219512939454,
+      "rewards/chosen": -0.37358812987804413,
+      "rewards/margins": 8.982958453042167,
+      "rewards/rejected": -9.356546606336321,
       "step": 40
     },
     {
-      "entropy": 1.1589677780866623,
-      "epoch": 4.103896103896104,
-      "grad_norm": 0.003875732421875,
+      "entropy": 1.1251012831926346,
+      "epoch": 2.050314465408805,
+      "grad_norm": 0.00433349609375,
       "learning_rate": 1.6666666666666667e-05,
-      "logits/chosen": -2.2044642238095227,
-      "logits/rejected": -1.7255148326997178,
-      "logps/chosen": -153.97796630859375,
-      "logps/rejected": -309.2699508666992,
-      "loss": 0.0002002128167077899,
-      "mean_token_accuracy": 0.5704852566123009,
-      "num_tokens": 310049.0,
+      "logits/chosen": -2.386099018384776,
+      "logits/rejected": -1.8694206899386892,
+      "logps/chosen": -120.46521186828613,
+      "logps/rejected": -271.48500633239746,
+      "loss": 0.0002081924758385867,
+      "mean_token_accuracy": 0.6062684953212738,
+      "num_tokens": 266060.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": -0.3379343111300841,
-      "rewards/margins": 10.036688327789307,
-      "rewards/rejected": -10.374622523784637,
+      "rewards/chosen": 0.5738538741134107,
+      "rewards/margins": 9.70900011062622,
+      "rewards/rejected": -9.13514620065689,
       "step": 41
     },
     {
-      "entropy": 1.1579881459474564,
-      "epoch": 4.207792207792208,
-      "grad_norm": 0.002685546875,
+      "entropy": 1.1459600627422333,
+      "epoch": 2.10062893081761,
+      "grad_norm": 0.0054931640625,
       "learning_rate": 1.5833333333333333e-05,
-      "logits/chosen": -2.295833219188626,
-      "logits/rejected": -1.7928715027581368,
-      "logps/chosen": -160.39765167236328,
-      "logps/rejected": -318.446325302124,
-      "loss": 0.00013650115579366684,
-      "mean_token_accuracy": 0.5852783843874931,
-      "num_tokens": 317868.0,
+      "logits/chosen": -2.3829266191138174,
+      "logits/rejected": -2.0208899907551676,
+      "logps/chosen": -131.44946670532227,
+      "logps/rejected": -271.2663917541504,
+      "loss": 0.00030793892801739275,
+      "mean_token_accuracy": 0.6211965084075928,
+      "num_tokens": 272769.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": -0.28954305313527584,
-      "rewards/margins": 10.830901622772217,
-      "rewards/rejected": -11.120444595813751,
+      "rewards/chosen": 0.10255242511630058,
+      "rewards/margins": 9.451286256313324,
+      "rewards/rejected": -9.34873378276825,
       "step": 42
     },
     {
-      "entropy": 1.1567590683698654,
-      "epoch": 4.311688311688312,
-      "grad_norm": 0.003326416015625,
+      "entropy": 1.0864861086010933,
+      "epoch": 2.150943396226415,
+      "grad_norm": 0.0211181640625,
       "learning_rate": 1.5e-05,
-      "logits/chosen": -2.4376148204179304,
-      "logits/rejected": -1.8646480940962602,
-      "logps/chosen": -138.2414083480835,
-      "logps/rejected": -304.6367073059082,
-      "loss": 0.0001619597605895251,
-      "mean_token_accuracy": 0.5863501876592636,
-      "num_tokens": 325576.0,
+      "logits/chosen": -2.313994988239532,
+      "logits/rejected": -1.9055890774453226,
+      "logps/chosen": -140.04273986816406,
+      "logps/rejected": -295.6635856628418,
+      "loss": 0.0007015014416538179,
+      "mean_token_accuracy": 0.623003862798214,
+      "num_tokens": 279603.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": -0.11933107685763389,
-      "rewards/margins": 10.622458934783936,
-      "rewards/rejected": -10.741789996623993,
+      "rewards/chosen": -0.03781823627650738,
+      "rewards/margins": 10.340316832065582,
+      "rewards/rejected": -10.37813514471054,
       "step": 43
     },
     {
-      "entropy": 1.1223484426736832,
-      "epoch": 4.415584415584416,
-      "grad_norm": 0.0019378662109375,
+      "entropy": 1.119643673300743,
+      "epoch": 2.20125786163522,
+      "grad_norm": 0.004425048828125,
       "learning_rate": 1.4166666666666668e-05,
-      "logits/chosen": -2.4513975092431988,
-      "logits/rejected": -2.025179191831654,
-      "logps/chosen": -116.16583919525146,
-      "logps/rejected": -285.6050682067871,
-      "loss": 9.024493920151144e-05,
-      "mean_token_accuracy": 0.6130669489502907,
-      "num_tokens": 333071.0,
+      "logits/chosen": -2.3158235133601206,
+      "logits/rejected": -1.804865597293398,
+      "logps/chosen": -136.55340480804443,
+      "logps/rejected": -278.1607780456543,
+      "loss": 0.00020963302813470364,
+      "mean_token_accuracy": 0.6163425669074059,
+      "num_tokens": 286436.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.3361262381076813,
-      "rewards/margins": 9.84731376171112,
-      "rewards/rejected": -9.511187613010406,
+      "rewards/chosen": 0.15877847000956535,
+      "rewards/margins": 9.487533926963806,
+      "rewards/rejected": -9.328755378723145,
       "step": 44
     },
     {
-      "entropy": 1.1869630813598633,
-      "epoch": 4.51948051948052,
-      "grad_norm": 0.005157470703125,
+      "entropy": 1.0829864144325256,
+      "epoch": 2.251572327044025,
+      "grad_norm": 0.006805419921875,
       "learning_rate": 1.3333333333333333e-05,
-      "logits/chosen": -2.1769293477016354,
-      "logits/rejected": -1.4691006673227756,
-      "logps/chosen": -158.22007369995117,
-      "logps/rejected": -318.19568634033203,
-      "loss": 0.0002636132121551782,
-      "mean_token_accuracy": 0.5729746297001839,
-      "num_tokens": 341054.0,
+      "logits/chosen": -2.4553567046643434,
+      "logits/rejected": -1.9043197747016694,
+      "logps/chosen": -137.17022514343262,
+      "logps/rejected": -297.57119941711426,
+      "loss": 0.00036281984648667276,
+      "mean_token_accuracy": 0.6070330739021301,
+      "num_tokens": 294223.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": -0.34203997533768415,
-      "rewards/margins": 10.176166713237762,
-      "rewards/rejected": -10.518206655979156,
+      "rewards/chosen": -0.006409928202629089,
+      "rewards/margins": 10.255102336406708,
+      "rewards/rejected": -10.261512219905853,
       "step": 45
     },
     {
-      "entropy": 1.0994098782539368,
-      "epoch": 4.623376623376624,
-      "grad_norm": 0.015380859375,
+      "entropy": 1.044785775244236,
+      "epoch": 2.30188679245283,
+      "grad_norm": 0.003997802734375,
       "learning_rate": 1.25e-05,
-      "logits/chosen": -2.385909726689113,
-      "logits/rejected": -2.06952027102175,
-      "logps/chosen": -155.6819829940796,
-      "logps/rejected": -289.1867733001709,
-      "loss": 0.0008207751670852304,
-      "mean_token_accuracy": 0.592678390443325,
-      "num_tokens": 349010.0,
+      "logits/chosen": -2.341592993550264,
+      "logits/rejected": -2.0601840101843556,
+      "logps/chosen": -113.64177894592285,
+      "logps/rejected": -268.7923049926758,
+      "loss": 0.0001799453020794317,
+      "mean_token_accuracy": 0.6692526936531067,
+      "num_tokens": 299634.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": -0.43019715102855116,
-      "rewards/margins": 9.629965007305145,
-      "rewards/rejected": -10.060162127017975,
+      "rewards/chosen": 1.3361206632107496,
+      "rewards/margins": 10.512819230556488,
+      "rewards/rejected": -9.176698446273804,
       "step": 46
     },
     {
-      "entropy": 1.1601977199316025,
-      "epoch": 4.7272727272727275,
-      "grad_norm": 0.005950927734375,
+      "entropy": 1.1353266835212708,
+      "epoch": 2.352201257861635,
+      "grad_norm": 0.0035400390625,
       "learning_rate": 1.1666666666666668e-05,
-      "logits/chosen": -2.420713169641715,
-      "logits/rejected": -1.8245980237599728,
-      "logps/chosen": -144.67862510681152,
-      "logps/rejected": -291.76464462280273,
-      "loss": 0.00025797143462114036,
-      "mean_token_accuracy": 0.5996529385447502,
-      "num_tokens": 356729.0,
+      "logits/chosen": -2.327246213935827,
+      "logits/rejected": -1.8727321034834783,
+      "logps/chosen": -161.3322401046753,
+      "logps/rejected": -302.8222484588623,
+      "loss": 0.00013208847667556256,
+      "mean_token_accuracy": 0.6015190705657005,
+      "num_tokens": 305920.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.29097576648928225,
-      "rewards/margins": 10.346692860126495,
-      "rewards/rejected": -10.05571711063385,
+      "rewards/chosen": 0.028157714928966016,
+      "rewards/margins": 10.57133662700653,
+      "rewards/rejected": -10.543178856372833,
       "step": 47
     },
     {
-      "entropy": 1.0812865048646927,
-      "epoch": 4.8311688311688314,
-      "grad_norm": 0.014892578125,
+      "entropy": 1.1054219380021095,
+      "epoch": 2.40251572327044,
+      "grad_norm": 0.006011962890625,
       "learning_rate": 1.0833333333333334e-05,
-      "logits/chosen": -2.4253862823366603,
-      "logits/rejected": -2.1636441064851186,
-      "logps/chosen": -130.41874885559082,
-      "logps/rejected": -258.45995330810547,
-      "loss": 0.0008855840424075723,
-      "mean_token_accuracy": 0.6128160208463669,
-      "num_tokens": 364748.0,
+      "logits/chosen": -2.3692881200690303,
+      "logits/rejected": -1.953993051539208,
+      "logps/chosen": -133.50615692138672,
+      "logps/rejected": -267.50397300720215,
+      "loss": 0.0003233029565308243,
+      "mean_token_accuracy": 0.6168623343110085,
+      "num_tokens": 313346.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.4050205182284117,
-      "rewards/margins": 9.163773000240326,
-      "rewards/rejected": -8.758752465248108,
+      "rewards/chosen": 0.12370538981485879,
+      "rewards/margins": 8.729556858539581,
+      "rewards/rejected": -8.605851471424103,
       "step": 48
     },
     {
-      "entropy": 1.1632622331380844,
-      "epoch": 4.935064935064935,
-      "grad_norm": 0.00103759765625,
+      "entropy": 1.0848723128437996,
+      "epoch": 2.452830188679245,
+      "grad_norm": 0.004974365234375,
       "learning_rate": 1e-05,
-      "logits/chosen": -2.427656775691583,
-      "logits/rejected": -1.7417183049324467,
-      "logps/chosen": -140.4052562713623,
-      "logps/rejected": -312.4628715515137,
-      "loss": 4.9687190767144784e-05,
-      "mean_token_accuracy": 0.5939344391226768,
-      "num_tokens": 372678.0,
+      "logits/chosen": -2.4114885162664828,
+      "logits/rejected": -1.933475836984313,
+      "logps/chosen": -138.67891216278076,
+      "logps/rejected": -284.93113708496094,
+      "loss": 0.0002774967870209366,
+      "mean_token_accuracy": 0.5793358683586121,
+      "num_tokens": 320662.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.09225330897606909,
-      "rewards/margins": 10.836278080940247,
-      "rewards/rejected": -10.744024753570557,
+      "rewards/chosen": -0.1479677117895335,
+      "rewards/margins": 9.403626084327698,
+      "rewards/rejected": -9.551593720912933,
       "step": 49
     },
     {
-      "entropy": 1.0998417615890503,
-      "epoch": 5.0,
-      "grad_norm": 0.0089111328125,
+      "entropy": 1.1292646825313568,
+      "epoch": 2.50314465408805,
+      "grad_norm": 0.0113525390625,
       "learning_rate": 9.166666666666666e-06,
-      "logits/chosen": -2.108919317236185,
-      "logits/rejected": -1.5721664481497553,
-      "logps/chosen": -105.09717407226563,
-      "logps/rejected": -249.9797332763672,
-      "loss": 0.00041876244358718395,
-      "mean_token_accuracy": 0.6374574661254883,
-      "num_tokens": 377580.0,
+      "logits/chosen": -2.3793244079557065,
+      "logits/rejected": -2.0578088532827836,
+      "logps/chosen": -151.90766048431396,
+      "logps/rejected": -276.3668155670166,
+      "loss": 0.0005854924093000591,
+      "mean_token_accuracy": 0.59138548001647,
+      "num_tokens": 326611.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.32345535755157473,
-      "rewards/margins": 8.785911655426025,
-      "rewards/rejected": -8.462456226348877,
+      "rewards/chosen": 0.3165478724986315,
+      "rewards/margins": 9.21978086233139,
+      "rewards/rejected": -8.903232991695404,
       "step": 50
     },
     {
-      "entropy": 1.1008698269724846,
-      "epoch": 5.103896103896104,
-      "grad_norm": 0.0166015625,
+      "entropy": 1.1128623560070992,
+      "epoch": 2.5534591194968552,
+      "grad_norm": 0.00640869140625,
       "learning_rate": 8.333333333333334e-06,
-      "logits/chosen": -2.1886847202899142,
-      "logits/rejected": -1.709530236018181,
-      "logps/chosen": -142.05830669403076,
-      "logps/rejected": -258.09220123291016,
-      "loss": 0.0009700299706310034,
-      "mean_token_accuracy": 0.6027038618922234,
-      "num_tokens": 385524.0,
+      "logits/chosen": -2.3023183294428042,
+      "logits/rejected": -1.967437557097666,
+      "logps/chosen": -116.14246654510498,
+      "logps/rejected": -265.47961044311523,
+      "loss": 0.000282328634057194,
+      "mean_token_accuracy": 0.6275674253702164,
+      "num_tokens": 332814.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": -0.39982338063418865,
-      "rewards/margins": 7.7835773229599,
-      "rewards/rejected": -8.183400690555573,
+      "rewards/chosen": 0.9145755774807185,
+      "rewards/margins": 9.993305206298828,
+      "rewards/rejected": -9.078729569911957,
       "step": 51
     },
     {
-      "entropy": 1.0808009058237076,
-      "epoch": 5.207792207792208,
-      "grad_norm": 0.0011444091796875,
+      "entropy": 1.1205245852470398,
+      "epoch": 2.6037735849056602,
+      "grad_norm": 0.001739501953125,
       "learning_rate": 7.5e-06,
-      "logits/chosen": -2.445027329506914,
-      "logits/rejected": -2.0271617790401937,
-      "logps/chosen": -128.9005184173584,
-      "logps/rejected": -312.5902729034424,
-      "loss": 5.263146886136383e-05,
-      "mean_token_accuracy": 0.6171352416276932,
-      "num_tokens": 393111.0,
+      "logits/chosen": -2.405199971394472,
+      "logits/rejected": -1.9401018785140567,
+      "logps/chosen": -120.2322769165039,
+      "logps/rejected": -278.4016227722168,
+      "loss": 9.171015699394047e-05,
+      "mean_token_accuracy": 0.611074797809124,
+      "num_tokens": 339547.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.2731950767338276,
-      "rewards/margins": 11.596651911735535,
-      "rewards/rejected": -11.323456883430481,
+      "rewards/chosen": 0.8479831870645285,
+      "rewards/margins": 10.514736771583557,
+      "rewards/rejected": -9.66675341129303,
       "step": 52
     },
     {
-      "entropy": 1.1334418207406998,
-      "epoch": 5.311688311688312,
-      "grad_norm": 0.006622314453125,
+      "entropy": 1.1072203814983368,
+      "epoch": 2.6540880503144653,
+      "grad_norm": 0.017822265625,
       "learning_rate": 6.666666666666667e-06,
-      "logits/chosen": -2.271390373833539,
-      "logits/rejected": -1.6907872117642864,
-      "logps/chosen": -141.5786190032959,
-      "logps/rejected": -274.7357540130615,
-      "loss": 0.00031432509422302246,
-      "mean_token_accuracy": 0.6188977956771851,
-      "num_tokens": 401012.0,
+      "logits/chosen": -2.2667942885965546,
+      "logits/rejected": -1.900450488756554,
+      "logps/chosen": -115.35403633117676,
+      "logps/rejected": -267.29714012145996,
+      "loss": 0.0008604128961451352,
+      "mean_token_accuracy": 0.5969742015004158,
+      "num_tokens": 346647.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": -0.02617559058126062,
-      "rewards/margins": 9.21671599149704,
-      "rewards/rejected": -9.242891550064087,
+      "rewards/chosen": 0.317084601148963,
+      "rewards/margins": 9.336723506450653,
+      "rewards/rejected": -9.019638895988464,
       "step": 53
     },
     {
-      "entropy": 1.2137573212385178,
-      "epoch": 5.415584415584416,
-      "grad_norm": 0.005767822265625,
+      "entropy": 1.1071247309446335,
+      "epoch": 2.7044025157232703,
+      "grad_norm": 0.0029754638671875,
       "learning_rate": 5.833333333333334e-06,
-      "logits/chosen": -2.392515574831131,
-      "logits/rejected": -1.7596199848269718,
-      "logps/chosen": -154.2357759475708,
-      "logps/rejected": -323.3644905090332,
-      "loss": 0.000225663956371136,
-      "mean_token_accuracy": 0.5656630396842957,
-      "num_tokens": 408880.0,
+      "logits/chosen": -2.324148264155996,
+      "logits/rejected": -1.7913556462377824,
+      "logps/chosen": -151.82556343078613,
+      "logps/rejected": -299.4059410095215,
+      "loss": 0.00012161008635303006,
+      "mean_token_accuracy": 0.6138368621468544,
+      "num_tokens": 353155.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": -0.2048426689580083,
-      "rewards/margins": 10.830295145511627,
-      "rewards/rejected": -11.035137951374054,
+      "rewards/chosen": 0.3809825163334608,
+      "rewards/margins": 10.480114579200745,
+      "rewards/rejected": -10.09913194179535,
       "step": 54
     },
     {
-      "entropy": 1.1423105299472809,
-      "epoch": 5.51948051948052,
-      "grad_norm": 0.00146484375,
+      "entropy": 1.16407360881567,
+      "epoch": 2.7547169811320753,
+      "grad_norm": 0.0017242431640625,
       "learning_rate": 5e-06,
-      "logits/chosen": -2.5769469207427296,
-      "logits/rejected": -2.021839826037388,
-      "logps/chosen": -133.6878776550293,
-      "logps/rejected": -303.7228317260742,
-      "loss": 7.118787470972165e-05,
-      "mean_token_accuracy": 0.5749763548374176,
-      "num_tokens": 416393.0,
+      "logits/chosen": -2.322006496808721,
+      "logits/rejected": -1.882509489706578,
+      "logps/chosen": -140.3326654434204,
+      "logps/rejected": -293.94667053222656,
+      "loss": 7.282369915628806e-05,
+      "mean_token_accuracy": 0.617231510579586,
+      "num_tokens": 358538.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.456505686044693,
-      "rewards/margins": 11.555195212364197,
-      "rewards/rejected": -11.098689556121826,
+      "rewards/chosen": 0.7881479188799858,
+      "rewards/margins": 10.648730993270874,
+      "rewards/rejected": -9.860582947731018,
       "step": 55
     },
     {
-      "entropy": 1.1627247333526611,
-      "epoch": 5.623376623376624,
-      "grad_norm": 0.003143310546875,
+      "entropy": 1.1090919971466064,
+      "epoch": 2.8050314465408803,
+      "grad_norm": 0.0311279296875,
       "learning_rate": 4.166666666666667e-06,
-      "logits/chosen": -2.370901666728044,
-      "logits/rejected": -1.86991563775274,
-      "logps/chosen": -124.0777416229248,
-      "logps/rejected": -270.50224685668945,
-      "loss": 0.0001636260567465797,
-      "mean_token_accuracy": 0.6245841830968857,
-      "num_tokens": 424280.0,
+      "logits/chosen": -2.31211133409438,
+      "logits/rejected": -1.8826621248132671,
+      "logps/chosen": -121.12357425689697,
+      "logps/rejected": -252.39348220825195,
+      "loss": 0.0018161134794354439,
+      "mean_token_accuracy": 0.6220331639051437,
+      "num_tokens": 364730.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.14732323656789958,
-      "rewards/margins": 8.864274621009827,
-      "rewards/rejected": -8.716951310634613,
+      "rewards/chosen": 0.17907649464905262,
+      "rewards/margins": 8.474798560142517,
+      "rewards/rejected": -8.295722126960754,
       "step": 56
     },
     {
-      "entropy": 1.167588233947754,
-      "epoch": 5.7272727272727275,
-      "grad_norm": 0.000553131103515625,
+      "entropy": 1.1059659719467163,
+      "epoch": 2.8553459119496853,
+      "grad_norm": 0.0093994140625,
       "learning_rate": 3.3333333333333333e-06,
-      "logits/chosen": -2.316611971228025,
-      "logits/rejected": -1.8084765653291615,
-      "logps/chosen": -154.08188915252686,
-      "logps/rejected": -332.9342460632324,
-      "loss": 2.3091437469702214e-05,
-      "mean_token_accuracy": 0.5768461301922798,
-      "num_tokens": 432075.0,
+      "logits/chosen": -2.1642121086590733,
+      "logits/rejected": -1.9605235069705975,
+      "logps/chosen": -142.26798820495605,
+      "logps/rejected": -266.46795082092285,
+      "loss": 0.0004599351086653769,
+      "mean_token_accuracy": 0.6000082641839981,
+      "num_tokens": 371209.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": 0.24338199198246002,
-      "rewards/margins": 12.503851771354675,
-      "rewards/rejected": -12.260469794273376,
+      "rewards/chosen": 0.3405567053705454,
+      "rewards/margins": 9.269334852695465,
+      "rewards/rejected": -8.928778231143951,
       "step": 57
     },
     {
-      "entropy": 1.1254538595676422,
-      "epoch": 5.8311688311688314,
-      "grad_norm": 0.01214599609375,
+      "entropy": 1.1505941599607468,
+      "epoch": 2.9056603773584904,
+      "grad_norm": 0.005279541015625,
       "learning_rate": 2.5e-06,
-      "logits/chosen": -2.1240879253748455,
-      "logits/rejected": -1.683307431601137,
-      "logps/chosen": -132.05009174346924,
-      "logps/rejected": -276.7969856262207,
-      "loss": 0.0006776420050300658,
-      "mean_token_accuracy": 0.6085913628339767,
-      "num_tokens": 439920.0,
+      "logits/chosen": -2.2861668483136004,
+      "logits/rejected": -1.8705101269968154,
+      "logps/chosen": -136.88138389587402,
+      "logps/rejected": -289.31751251220703,
+      "loss": 0.0002126125618815422,
+      "mean_token_accuracy": 0.6103544309735298,
+      "num_tokens": 378077.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": -0.0067318761721253395,
-      "rewards/margins": 8.990104794502258,
-      "rewards/rejected": -8.996836602687836,
+      "rewards/chosen": 0.38323727436363697,
+      "rewards/margins": 10.188703656196594,
+      "rewards/rejected": -9.805466413497925,
       "step": 58
     },
     {
-      "entropy": 1.1096209362149239,
-      "epoch": 5.935064935064935,
-      "grad_norm": 0.00762939453125,
+      "entropy": 1.0992570891976357,
+      "epoch": 2.9559748427672954,
+      "grad_norm": 0.00213623046875,
       "learning_rate": 1.6666666666666667e-06,
-      "logits/chosen": -2.464917742945188,
-      "logits/rejected": -2.0316993109334662,
-      "logps/chosen": -165.6044044494629,
-      "logps/rejected": -321.6104621887207,
-      "loss": 0.0003184170927852392,
-      "mean_token_accuracy": 0.5715851411223412,
-      "num_tokens": 448088.0,
+      "logits/chosen": -2.298579599751336,
+      "logits/rejected": -2.0673505464671744,
+      "logps/chosen": -141.01217555999756,
+      "logps/rejected": -277.6583061218262,
+      "loss": 0.00010025675874203444,
+      "mean_token_accuracy": 0.6147632747888565,
+      "num_tokens": 384115.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": -0.5723045284976251,
-      "rewards/margins": 10.441723942756653,
-      "rewards/rejected": -11.01402860879898,
+      "rewards/chosen": 0.6212236363207921,
+      "rewards/margins": 10.357111930847168,
+      "rewards/rejected": -9.735888302326202,
       "step": 59
     },
     {
-      "entropy": 1.177198052406311,
-      "epoch": 6.0,
-      "grad_norm": 0.01043701171875,
+      "entropy": 1.094216755458287,
+      "epoch": 3.0,
+      "grad_norm": 0.02587890625,
       "learning_rate": 8.333333333333333e-07,
-      "logits/chosen": -2.246316011456704,
-      "logits/rejected": -1.7093484860379249,
-      "logps/chosen": -140.52530975341796,
-      "logps/rejected": -272.9529998779297,
-      "loss": 0.00047754126717336476,
-      "mean_token_accuracy": 0.584682285785675,
-      "num_tokens": 453096.0,
+      "logits/chosen": -2.3206561180335696,
+      "logits/rejected": -1.7881133994765854,
+      "logps/chosen": -137.9519805908203,
+      "logps/rejected": -263.429922921317,
+      "loss": 0.0014771231217309833,
+      "mean_token_accuracy": 0.610472559928894,
+      "num_tokens": 390000.0,
       "rewards/accuracies": 1.0,
-      "rewards/chosen": -0.20170196890830994,
-      "rewards/margins": 8.390394496917725,
-      "rewards/rejected": -8.592096328735352,
+      "rewards/chosen": -0.13421392613755806,
+      "rewards/margins": 8.44434186390468,
+      "rewards/rejected": -8.578555720193046,
       "step": 60
     }
   ],
   "logging_steps": 1,
   "max_steps": 60,
   "num_input_tokens_seen": 0,
-  "num_train_epochs": 6,
+  "num_train_epochs": 3,
   "save_steps": 60,
   "stateful_callbacks": {
     "TrainerControl": {
@@ -1107,7 +1107,7 @@
       "attributes": {}
     }
   },
-  "total_flos": 3689526234746880.0,
+  "total_flos": 3189182437748736.0,
   "train_batch_size": 1,
   "trial_name": null,
   "trial_params": null